Skip to content

Open research · Law × AI

Where law
meets code

Open infrastructure for verified legal work with AI. Working cards on the AI Act and GDPR, practical tools and open research questions, each with a record of how it was built and what in it has been verified.

Justitia, the statue of justice (blue duotone)
01 AI Act commentary

Working cards on the EU AI Act: risk classification, deadlines and an article-by-article reading.

02 GDPR commentary

The regulation article by article, a compliance guide and original document templates.

03 AI-literacy course

A ready training under Article 4 of the AI Act that anyone can take, adapt to their setting and build a product on.

04 Prompt builder

Assembling prompts for AI that hold perspective, sources and verification.

05 Research questions

Open questions at the border of law and AI, worked on as they go.

What is being built here

Legal To Code is open infrastructure for verified legal work with artificial intelligence. It collects research questions, practical tools and notes on how to translate legal work into a language machines understand while keeping it verifiable.

The content is produced in a legal workflow with agentic AI and draws on my own attempts, notes and collected papers, including other authors’ thinking. The workflow is governed by machine-enforced rules matching my methodology: a cited section is checked word for word against the library of legislation, and interpretive text is read article by article. I keep adding to and updating the site, and every change is listed with its date in the list of changes. The detail is in how this site is produced.

How content is made

Content is produced in the Legal To Code legal workflow for Claude Code (version 1.152.0). Cited provisions are verified against a local library of their verbatim wording. Where the wording is not in the library, it is verified in the Czech Collection of Laws. For case-law research the workflow is connected to the Salvia MCP server. Automated checks of citations and internal consistency are part of the workflow. Verify decisive information against the original sources.

How this works · What has changed